Research map / Reinforcement learning / Visual and representation RL
Universal Morphology Control via Contextual Modulation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add fixed attention (morphology-only attention) to MetaMorph* (ablation) | MetaMorph* · Unseen morphologies (topology), 5 environments | episodic return: FT 1266→1439; Incline 312→468; Exploration 19→22; VT 767→860; Obstacles 829→937 | Helped |
| add hypernetworks (context-conditioned embedding and decoder) to MetaMorph* (ablation) | MetaMorph* · Unseen morphologies (topology), 5 environments | episodic return: FT 1266→1259; Incline 312→312; Exploration 19→16; VT 767→900; Obstacles 829→969 | Mixed |
| add both fixed attention and hypernetworks to MetaMorph* | MetaMorph* · Unseen morphologies (topology), 5 environments | episodic return: FT 1266→1490; Incline 312→403; Exploration 19→23; VT 767→971; Obstacles 829→1133 | Helped |
| add hypernetworks to MetaMorph* in Exploration (ablation) | MetaMorph* · Exploration training | training return | Hurt |
| add height map as input to attention computation (terrain-conditioned attention) (ablation) | FA variant · Variable terrain training | training return | Hurt |
| remove dropout from MetaMorph (inconsistent dropout masks removed) (ablation) | MetaMorph · FT training | training return | Hurt |
| add learned positional encoding to MetaMorph node embeddings (ablation) | MetaMorph · Multi-robot training (FT) | training return | No clear effect |
| contextual modulation (full ModuMorph) vs MetaMorph* on multi-robot training | MetaMorph* · Training morphologies, 5 environments | final return improvement: 19%, 53%, 48%, 31%, 29% in FT, Incline, Exploration, VT, Obstacles | Helped |
Closest papers by meaning
- Distilling Morphology-Conditioned Hypernetworks for Efficient Universal Morphology Control (ICML 2024 · similarity 0.92)
- DittoGym: Learning to Control Soft Shape-Shifting Robots (ICLR 2024 · similarity 0.89)
- Any2Policy: Learning Visuomotor Policy with Any-Modality (NeurIPS 2024 · similarity 0.86)
- Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformers (NeurIPS 2024 · similarity 0.86)
- Convergent Functions, Divergent Forms (NeurIPS 2025 · similarity 0.86)
- MorphVAE: Advancing Morphological Design of Voxel-Based Soft Robots with Variational Autoencoders (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.